Inspiration

Shufflow started with our cofounder Tianhui’s experience with ADHD. She often found it difficult to stay focused on one task until it was finished. When she became stuck, switching to another related task helped her keep working and make progress.

This made us question how most productivity tools are designed. Many of them expect users to choose one task, stay focused, finish it, and then move to the next one. But people do not always work this way.

During a normal day, people may be interrupted, lose focus, get stuck, or have more energy for certain tasks than others. The difficult part is often starting a task, deciding what to do next, switching when a task is not going well, or returning to unfinished work. We call these points transition moments.

Most productivity tools help users list and organize their tasks, but offer little support during these transition moments.

To understand whether other people had the same problem, we conducted two rounds of research with 15 adults. Some had been diagnosed with ADHD, while others reported similar attention-related difficulties. Three common problems came up:

Starting a task
Restarting after a distraction
Remembering where they left off in a multi-step task

These findings led us to create Shufflow: a productivity app that helps people continue making progress even when their focus changes during the day.

What it does

Shufflow uses AI to turn complex work into connected, manageable steps while preserving task context.

When a user becomes distracted or stuck, Shufflow can suggest another relevant action based on factors such as available time, current energy, priority, and difficulty. The user can continue making meaningful progress instead of abandoning the work session.

When the user returns to an earlier task, Shufflow restores its context so they can resume without reconstructing what they were doing.

Shufflow also includes:

AI-assisted task breakdown
AI-assisted, context-aware task recommendations
Summaries of completed and upcoming work
Time-aware chimes that support time awareness
A clean, low-distraction interface
A new fidget-inspired interaction system created during OpenAI Build Week

Existing before Build Week

The following existed before the submission period:

The original Shufflow product concept
An AI-assisted task-breakdown workflow
Task-recommendation and prioritization logic
The context-preservation concept for interrupted tasks
An initial web prototype

What we built during OpenAI Build Week

Shufflow is a pre-existing project that began on September 20, 2025. The original prototype, preliminary user research, and baseline task-management workflow existed before the Build Week submission period.

For OpenAI Build Week, we used GPT-5.6 and Codex to create a substantially new fidget-inspired interaction layer. It was inspired by an insight from our user interviews: many people with attention difficulties use physical fidget objects as part of their daily routines.

Rather than simply reproducing the appearance of a fidget toy, we studied how people interact with them to create a new interaction language for Shufflow:

Simple movements
Repeated actions
Minimal effort
Immediate visual feedback

What we added during Build Week

Flow Cube

We created the Flow Cube as the main visual symbol of Shufflow’s workflow. It shows that work can move, pause, and continue without losing its place. Users interact with the cube to move between tasks while staying aware of their current progress.

Swipe-based task switching

Users swipe right to move to a recommended next task and swipe left to return to the previous task. The gestures are quick and simple, but every switch is still chosen by the user.

Adaptive timer and visual feedback

We redesigned the timer and visual feedback around the new interaction system. Important information remains visible, while each action produces a clear response without adding unnecessary distraction.

How we built it

We started with an insight from our user interviews: several participants with attention difficulties regularly used physical fidget toys.

We used GPT-5.6 to:

Research and group common fidget interactions
Understand why simple, repeatable movements may help users stay engaged
Explore how these movements could work in a digital product
Connect them to moments when a user starts, switches, or returns to a task
Turn the selected ideas into clear product requirements

Our team reviewed these ideas and decided that the fidget toy should not become a separate feature inside Shufflow. Instead, its simple movements and immediate feedback would become an interaction language used throughout the app.

We then worked with Codex to turn the product requirements into implementation tasks. Codex helped us build:

The Flow Cube and its visual states
Swipe-right and swipe-left task transitions
Connections between the new interactions and the existing task state
The task-context recovery experience
The redesigned timer and visual feedback

We also used Codex to understand the existing codebase, find implementation problems, fix state-management issues, and refine the gestures and interface behavior. Our team reviewed the changes and made the final product, design, and technical decisions.

At each stage, we asked the same question:

Does this interaction make it easier to start, switch, or return to a task without creating another distraction?

GPT-5.6 helped us research, compare, and organize possible approaches. Codex helped us build and improve the selected approach. Together, they helped us move more quickly from a user-interview insight to a working feature that people can try.

Challenges we ran into

Our main challenge was turning ideas from physical fidget toys into digital interactions that people could understand and use easily.

Physical fidget toys are simple, and immediate. On a screen, the same idea can quickly become a decoration or another source of distraction.

We had to balance three needs:

The interaction should be engaging enough to help the user take action.
It should require very little thought or learning.
It should keep the interface calm and free from unnecessary distractions.

This balance affected many small design decisions. We refined the design by removing unnecessary movement, keeping the gestures consistent, and ensuring that every interaction had a clear purpose: helping the user start, switch, or return to a task.

To address the challenge, we worked in quick user-feedback cycles:

Show users the latest version.
Ask what felt clear, confusing, helpful, or distracting.
Make changes based on their feedback.
Test the revised version again.

This process helped us improve the app quickly. It also helped us verify that the new interactions supported users without creating more distraction.

What we learned

Building the fidget interaction system taught us that a visual interaction cannot be added as an isolated animation. Every movement needs to represent a real action in the product and connect correctly with the rest of the workflow.

For each interaction, we had to define:

What the user is trying to do
What the app’s current state is
What changes after the interaction
What information needs to be saved
How the user can return or correct the action
What visual feedback confirms that the action worked

We also learned the value of building in short cycles. We implemented one interaction, tested it within the full workflow, collected feedback, and refined it before moving on.

What’s next for Shufflow

We are onboarding early testers and plan to develop a user-ready version over the next two months.

Our next research cycle will specifically evaluate the new fidget-inspired interaction system. We plan to measure whether it:

Engage user to get started
Reduces the time required to choose a next action
Helps users return to interrupted work

Longer term, we want Shufflow to become a flexible productivity tool that adapts to different patterns of attention without judging or penalizing users for how they work.